The Research On Image Noise Suppression Of CMOS Image Sensor

With the development of manufacturing technology of Integrated Circuit and the improvement of IC design along with its advantages, CMOS Image Sensor using CMOS techniques have been widely used in many areas because of its merits, such as high integration density, low power dissipation, minute extension, simple process, low-cost and short development cycle. The market is wide open, especially in the area of digital products, therefore it is very useful and valuable to research and develop the CMOS Image Sensors.At first, in this paper, an overview of history and status about CMOS image sensors was presented sketchily home and abroad. After having compared CMOS image sensor with CCD image sensor on performance characteristics, the purpose and significance of this subject study was pointed out. The paper described briefly the structure and working principle of CMOS image sensor, as well as its three types of pixel unit structure. It also gave a detailed account of diverse noises and their sources existed in its pixel unit, on the ground of analyzing the readout circuit of CMOS image sensor. Meanwhile, noise model was built, and its corresponding suppression technologies provided. The focal point of this paper is about how to suppress another type of noise, Fixed-Pattern-Noise (FPN), which exists in column readout circuit of CMOS image sensor. So the discussion about FPN was done, and a new kind of noise suppression circuit put forward too. Finally, having been tested in simulated environment HSPICE, the new type of circuit proved effectively.In order to capture better quality picture and minimize the noise effect farther, it's necessary to deal with the image collected by CMOS image sensor at the back-end. The paper proposed an adaptive filtering method to remove digital noises. During the process, two filters which include median filter and Gaussian filter were designed through FPGA. At last, according to the method, the image, which was collected by CMOS image sensor at back-end, was processed in simulated environment MATLAB.